Recherche avancée

Sur d’autres sites (4127)

• Debugging in Visual Studio without Pdb files (C++ Access Voilation)

  2 décembre 2016, par Prakash M

  I built OpenCV binaries(.dll) using Cmake & visual studio which generated .pdb file which helped me to find the issue in code (Partially !)

  How this Crash is being caused. .
  I’m using a software with which we can set internet download speed limit (transfer rate) for any particular program.

  Now if i connect IP camera to the code below, i noticed that my app needs around 100Kb/s of internet usage (transfer rate) - only then i can watch live stream seamlessly.
  Lets say i cut down (set) my application internet usage to 10Kb/s [This is the reason behind crash]
  in this case, i should be able to see a new frame once in 4+ seconds.

  I’m getting access violation error probably because (cap>>img;) cap is trying to reach a location in ram & get the frame but there is no frame YET because its still being downloaded due to low internet speed.
  Clearly the pointer is reaching some location in ram to grab a frame which is not yet present.

  Some interesting behaviour . . .

  Void OpenCamera()
  
  {
  
      VideoCapture cap("http://192.168.1.3:8080/video?x.xmjpeg");
  
      Mat img;
  
      while(true)
  
      {
  
        try
  
        {
  
          if(cap.isOpened()) //also tried grab + retrieve, crashes at grab
  
          cap>>img; //code crashes here
  
        }
  
        catch(...)
  
        {
  
          cout<<"Camera Disconnected"<code>

  If i use the entire code in same class (within same header file), there is no problem at all(new frame is displayed after 4+ seconds without crashing the program) but if i put the code into a separate class(different header file), then call the function to open camera from a class object, then it crashes if internet speed is cut down.
  weird behavior - if i debug step by step, it never crashes !
  

  when i build opencv library with ffmpeg , i get .pdb file only for opencv (opencv_world310.pdb)- so no issue debugging using call stack
  but i do not get pdb for ffmpeg (because Opencv_ffmpeg.dll is precompiled and that is where its crashing)

  hence its getting hard to debug, building ffmpeg doesn’t produce pdb file cause its built using MSYS
  so is it possible to debug with what we have ?

  I’m including snapshot from visual studio debugging,
  some of the variables that will help in understanding :

  typedef int (*CvGrabFrame_Plugin)( void* capture_handle );      [cap_ffmpeg_api.cpp]
  
  protected: void* ffmpegCapture;                                 [cap_ffmpeg.cpp]
  
  static CvGrabFrame_Plugin icvGrabFrame_FFMPEG_p = 0;            [cap_ffmpeg.cpp]

  

  Exception thrown at 0x0A0AF6F0 (opencv_ffmpeg310.dll) in Sample.exe :
  0xC0000005 : Access violation reading location 0x00000020. If there is
  a handler for this exception, the program may be safely continued.

  in source code i included below line & compiled & used it in project - didn’t work, crashed again !
  if(ffmpegCapture) - null pointer check

  can we make some changes at line 214 in [cap_ffmpeg.cpp] to avoid crash ?
  other header files are just one folder up.

  Update : I noticed that program crashes immediately when i limit internet consumption speed. I’m using C++/Cli(winforms, target dot net Framework = 4.6), i have CameraClass (in separate header file) & main function in (separate header file)
  Main function has below code
  

  CameraClass ^CC = gcnew CameraClass();
  
  CC->OpenCamera();

  Some clash between .net memory handling & C++ memory handling ?

• Merging .flv chunks from stream site

  4 décembre 2016, par Iztooi

  I’m trying to download a video from an alternative version of the Ustream site. This site is apparently used for a DRM protection, in fact you can’t watch this video without having access to it through a login from this site.
  The video system should be the same used by ustream (same links), that is calling different flv chunks every x seconds and merging them into the stream (similar to an rtmp service, but it doesn’t use this protocol).

  I managed to download every .flv chunks composing this video, but i had issues on merging them into a single file. First of all these flvs files have some audio/video lags/problems, but since there is no other way to download this video i’m not complaining about lags.

  The real issue is that every flv file has an additive increase on the duration, despite having a normal size. Video duration is about 6 seconds, so the first chunk is 6 seconds long, the second chunk is 12 seconds long, although only the first 6 are visible (the remaining 6 are "corrupted"). If i try to merge (with ffmpeg or any video-joining software) all these files (about 600-700, for a total duration of 1 hour and 15 minutes) only the first two sections are visible, then I cannot see anything. (Being more clear, if i try to merge the first 2 chunks, file is 18 seconds long and i can see the first 12 ; if i try to merge the first 3 chunks, file is 36 seconds long and i can still only see the first 12 seconds.)

  Anyone has any idea ?

• Generate individual HLS-compatible .ts segments on-demand by downloading as little bytes as possible from a remote input file

  27 janvier 2017, par Romain Cointepas

  I’m trying to generate individual HLS-compatible .ts segments on-demand by downloading/reading as little bytes as possible from a remote input file (hosted on a server supporting byte-ranges requests).

  One of the application for this would be to be able to transcode and play on Apple TV (via Airplay) a remote file that is not Airplay compatible, without having to download the entire file first.

  I am generating the playlist myself, and I have access to the ffprobe results for the remote file (that gives video duration, etc.).

  I have something working that plays via Airplay but with small video and audio glitches between each segments when I use the following command to generate each segment :

  ffmpeg -ss 60 -t 6 -i http://s3.amazonaws.com/misc-12345/avicii.vob -f mpegts -map 0:v:0 -map 0:a:0 -c:v libx264 -bsf:v h264_mp4toannexb -force_key_frames "expr:gte(t,n_forced*6)" -forced-idr 1 -pix_fmt yuv420p -colorspace bt709 -c:a aac -async 1 -preset ultrafast pipe:1

  Note : above command is for segment 11.ts, and in the m3u8 playlist I advertise each segment duration as 6 seconds.

  Here is a Youtube video showing the audio/video glitches between segments :
  https://www.youtube.com/watch?v=0vMwgbSfsu0

  The segment or hls modules of ffmpeg can’t be used because they both generate all the segments at once.

  I’ve been struggling on this for some days now and I would really appreciate some help !